Tuesday, March 24, 2026

Start of archive·May 16

Top story

The NumbersTechCrunch Venture

With $3.5B in fresh capital, Kleiner Perkins is going all in on AI

One of Silicon Valley's most prestigious VCs is making its biggest bet ever on AI, signaling massive capital deployment across both early and late-stage companies in the space.

$3.5 billion total fundraise

The briefs

Alibaba's RISC-V server chip represents a critical shift toward domestic semiconductor independence for China's AI stack, reducing reliance on US-controlled architectures and enabling optimized deployment of local LLMs at scale.

Alibaba DAMO Xuantie C950 — RISC-V based server chip

Arm's move from IP licenser to chip manufacturer signals a fundamental shift in AI infrastructure strategy—vertical integration is now table stakes for compute advantage. This directly threatens NVIDIA's moat and reshapes the supply chain for frontier AI labs.

Arm entering chip manufacturing (strategic pivot from design-only model)

Government regulatory action against a leading AI lab raises questions about supply-chain designations and their use as competitive tools. This sets a precedent for how national security apparatus may constrain AI development.

District court judge questioned DoD motivations for labeling Anthropic a supply-chain risk

Anthropic is shipping agentic capabilities (computer control) as a product feature, signaling the market transition from conversational AI to autonomous agents. This is a direct competitive move against similar launches from OpenAI and others in the agent-layer race.

Claude gains computer control capability

Vertical integration of AI compute across three capital-intensive companies signals that inference and training bottlenecks are no longer a software problem—they're a hardware supply problem. This moves the needle on who controls the AI stack.

Chip megaproject spans Tesla, SpaceX, and xAI

Major VCs are betting big on India's emerging AI talent pool, with highly selective programs targeting early-stage innovation outside traditional tech hubs.

6 startups selected

Apple's latent lookahead training addresses a core limitation of next-token prediction: the inability to reflect on alternative paths. This could shift how reasoning and inference efficiency are architected in future LLMs, competing directly with other approaches to improve model expressiveness without proportional compute increases.

Accepted at ICLR 2026 Workshop on Latent & Implicit Thinking

Mozilla's cq positions itself as a knowledge repository and coordination platform purpose-built for AI agents, addressing a gap in how autonomous systems discover, share, and execute multi-step tasks. This matters because it signals how the agent economy is moving from capability to tooling — the picks-and-shovels layer that makes agents actually deployable at scale.

Mozilla launches cq product

Microsoft is embedding generative AI deeper into its productivity suite, turning OneDrive into a creative tool. This signals a shift from 'AI as add-on' to 'AI as core feature' — and a direct play to lock users into the Microsoft ecosystem.

Microsoft ships AI photo restyling feature to OneDrive